In 2000, Nortel Networks represented an ‘index-distorting’ one-third of the TSE 300 Index, creating undue risk for many funds. Some funds managed this risk by underweighting Nortel, which created active risk relative to their policy portfolio. Others chose a policy response, adopting a capped index that limited how much could be invested in a single stock. Both the active management and policy responses were band-aid solutions that treated the symptom of undue risk, but not the problem. The underlying problem was that asset-based processes—those that set target weights and minimum/maximum position limits, although simple, had become less effective in dealing with today’s complicated and evolving portfolios.

A more effective and timely way to deal with the Nortel problem would have been to adopt a riskbased approach to portfolio management. Such an approach would have involved measuring risk more frequently, setting limits on risk and rebalancing the portfolio based on risk/return assessments rather than asset mix targets. In its 2001 annual report, the Canada Pension Plan Investment Board described how it avoided more than $500 million in potential losses, on average assets of $5 billion, related to Nortel through a “risk management initiative” that moved the fund from passive to “partially active” investing. It’s remarkable how uncommon this approach was at the time, and still is, but that’s about to change.

Risk budgeting is the process of allocating risk in an explicit way. Like all budgeting processes, it allocates a scarce resource(risk)to meet an objective(maximize returns). It has the same goal as asset-based processes, but that’s where the similarities end. In risk budgeting, the focus is on risk and return, and the asset mix is a byproduct. For asset-based processes, it’s the other way around—the focus is on assets and returns.

Companies that define their processes in terms of “why” (objectives)rather than “how”(means)are more likely to evolve with changing times. That’s why XEROX calls itself a document management company— not a company that makes photocopiers. It’s also why risk budgeting is better than asset-based processes. Risk budgeting acknowledges that a constant asset mix has a changing risk profile (as the Nortel example illustrates)and that rebalancing should be based on risk and return assessments, rather than asset weights.

Another costly asset-based constraint is the “long-only” constraint, which imposes a minimum(0%)and maximum(100%) allocation to assets. The constraint is designed to mitigate potentially large losses from short selling. The large cost of the constraint, which is widely acknowledged, could be reduced if risks were controlled directly using risk-based limits.

The popular 50% currency hedge ratio represents a further asset-based constraint that may impose a cost. If hedge ratios are limited to be between 0% and 100%, the selection of 50% is convenient for those who want to maximize the room for active management. Unfortunately, having the same hedge ratio for all currencies may not be optimal in the long term. It may be the case—depending on a fund’s liabilities— that certain funds should adopt different hedge ratios for different currencies, just as they have different asset allocations for different asset classes.

Asset-based processes have one redeeming quality. They’re simple. A typical process might involve finding the asset mix that meets a return objective, while minimizing surplus at risk(the risk that assets rise less than liabilities). Unfortunately, the process is too simple. The surplus at risk and active risks that flow from this process might be discussed, but these risks are not reviewed and updated regularly to reflect changes in market conditions. As a result, those who manage and oversee the fund are less likely to appreciate the dynamics of the risks that the fund takes through time. The greater focus on returns (rather than risk)provides a false sense of comfort—especially when you consider that returns are less predictable than risks. When bad things happen, as they surely will, a frequent reaction is one of shock— at either the severity or frequency with which certain losses occur.

Risk-based processes are harder to understand, but are more effective and will result in higher risk-adjusted returns(more efficient portfolios)because better(risk) measurement will lead to better(risk)management. Processes that rebalance based on risk assessments relative to return expectations will avoid undue risk more often than processes that rebalance based on fixed asset targets.

To implement risk budgeting in practice, pension funds need to answer at least five questions:

1. What risks should we manage?

2. How much return do we need for risks that we take?

3. How much risk is too much?

4. Where should we take risk?

5. Did we get paid enough for the risks we took?

The answers to these questions could flow from a risk management framework that provides a link between a fund’s investment strategy and its mission, values and beliefs. Such a risk framework should include at least five elements:

• A minimum risk portfolio(MRP)that becomes the primary benchmark for assessing risk and performance.

• A cost of risk capital that acknowledges risk as a scarce resource that has a cost(risk premium), and that a higher return is required for activities that involve incrementally higher risk.

Risk limit(s) that replace, or at least supplement, asset-based limits that are too costly and inconsistent in their treatment of different portfolios over time.

• A risk budget that allocates the target level of risk to various assets and managers to produce a required return in the most efficient way possible.

• An assessment process that measures economic value added that takes into account returns and risk on a regular basis(i.e. monthly or quarterly).

The MRP is the policy portfolio if risk budgeting is applied to active management. For surplus management(“liability-driven investing”), the MRP is the liabilities, as represented by a portfolio of securities(mostly fixed income).

The cost of risk capital might depend on the equity risk premium, in the case of surplus management. In an active management context, the cost of risk capital at the total portfolio level would depend on the number of active programs and their size.

A starting point for developing risk limits is to calculate what surplus and active risks are implied by the current portfolio. In other words, calculate what the maximum surplus and active risk would be using the asset-based policy targets and minimum/ maximum limits as constraints. A typical 60/40 asset mix, for example, might have a surplus at risk of 11%—one year in 10 or 10% of the time, assets might be expected to grow less than liabilities by 11% or more. One year in 100 or 1% of the time, assets might underperform liabilities by 22% or more. The active risk limits implied for the total portfolio would depend on the range of the minimum/maximum bands for asset classes and the extent to which active management is pursued within asset classes and other activities.

Funds that applied risk budgeting processes in the past few years were probably less surprised(or not at all surprised)by the severity and frequency of losses during the recent “perfect(pension)storm”, where assets did poorly and liabilities rose with falling real interest rates. Such funds were likely better prepared for the storm and may have decided to react differently once the storm subsided.

The risk budget is the risk-based equivalent of a target allocation, except that the allocation is expressed in terms of risk rather than assets. This can be presented in many ways but the most informative way is to measure and compare the impact of small changes in asset allocations on both risk and return. A risk budget presented on a surplus basis might have shown Nortel contributing a great deal of risk in 2000—more than could be justified by any reasonable return expectation for the stock. That risk budget might also have shown that real return bonds looked very attractive because they reduced risk much more than they reduced expected returns, given their low allocation in most portfolios.

The assessment process involves measuring risk as frequently as returns. Ideally, someone is held accountable for this performance by linking risk-adjusted performance to compensation. This is easier said than done, and it’s easier to do in an active management context, where correlations are low, than surplus management(“liability-driven investing” or LDI), where correlations are higher. Why is it harder for LDI? Because LDI is the ultimate team sport— where individual specialists(fixed income)may be asked to play a total portfolio game.

Canada’s two largest DB plans, the Ontario Teachers’ Pension Plan and the Canada Pension Plan Investment Board, have applied these frameworks and processes in managing both surplus and active risk. But is it time for smaller plans to implement risk-based processes? Maybe. The benefit of doing so increases over time as portfolios become more complex. In 1990, the average Canadian DB plan had two-thirds of its assets in fixed income. Portfolio management was easier because two asset classes(Canadian stocks and bonds)represented 90% of the portfolio. Today, half of the bonds(about one-third of the total portfolio)is in other, more complex, assets—a fact that makes risk measurement much harder.

Fortunately, the costs of maintaining the risk systems and data needed to implement risk budgeting in practice are falling. They’re still big, but much smaller than the cost associated with undue risk— risk that is higher than it has to be or risk that is not well understood and which could lead to poor portfolio choices.

We may have reached the point where smaller funds have the comparative advantage over larger funds when it comes to risk budgeting. When the Canada Pension Plan avoided large losses on Nortel in 2000/01, its investment department consisted of two people—the chief executive officer and its vice-president of research and risk management. Today, this fund—with closer to two hundred people than two— would have more resources at its disposal than in 2000, but it would have other challenges to overcome if it were to implement risk budgeting from scratch. Portfolio management is a team sport and no paradigm shift is ever easy, but having fewer cats to herd makes it easier.

Valter Viola is president of Holland Park Risk Management Inc. in Toronto.

For a PDF version of this article, click here.